On Tue, May 12, 2015 at 10:50 AM, Matthias Goldhoorn <[email protected]> wrote: > +1 by doing so we separate the HP generation from the API (which IS > currently a problem)
The main problem is that the package directory should point to the API directory. And we should *definitely* avoid broken links. Meaning, the package directory, when it gets generated, should know whether there *is* an API for a particular package. When I say that it can be solved, what needs to be done is to more cleanly separate the build from the website generation. One idea is to: - from a build, have a script *update* a set of JSON files with: - an index of the packages (name, URL, documentation URL, status of build) that gets rendered by a client-side JS page - an index of the oroGen tasks and types, which is also rendered client-side. This one should ideally be generated from an unpacked model-pack instead of directly from the build's install folder (would make updating it easier, and therefore separate the two). The JSON files would be uploaded to a separate domain (could be the current server, or even just a S3 bucket which would cost around 1 cent a month). Sylvain _______________________________________________ Rock-dev mailing list [email protected] http://www.dfki.de/mailman/cgi-bin/listinfo/rock-dev
